Load dependencies

Load imager

Show dimension of an image

plant<- load.image('~/Documents/Github/PhD//imageprocessing/burseragrav1.jpg')
dim(plant)
## [1] 1400 2100    1    3

Image

plot(plant)

## Y-axis shift

imshift(plant,0,50) %>% plot

## X-Axis

imshift(plant,50,0) %>% plot

Both Axis Shift

imshift(plant,50,50) %>% plot

Resize the image

map_il(2:4,~ imresize(plant,1/.)) %>% imappend("x") %>% plot

Rotate

imrotate(plant,30) %>% plot

Convert To Black & White

grayscale(plant) %>% plot

Mask Cut

msk <- px.flood(plant,400,900,sigma=.28) %>% as.cimg
inv<- msk %>%as.data.frame
inv$value<- as.numeric(inv$value==0)
msk<- inv %>% as.cimg()
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values
plot(msk*plant)

Filter out data using a set of pixels based on either EBG or HSV values

subsetImage<- function(test,xrange,yrange){
  newtest<- test %>% as.data.frame
  dfnew<-subset(newtest, x%in%xrange & y%in%yrange) 
  dfnew<- as.cimg(dfnew)
  return(dfnew)}

indicatorMatrix <-function( test,xrange,yrange,types)
{

if(types=='RBG'){
newtest<- test %>% as.data.frame} else{newtest<-RGBtoHSV(test) %>% as.data.frame}   
dfnew<-subset(newtest, x%in%xrange & y%in%yrange) %>% ddply( .(cc),summarize,SD = sd(value, na.rm=TRUE),Mean = mean(value, na.rm=TRUE))

tmp<-subset(newtest,cc==1)
tmp$value<-as.numeric( between (newtest$value[newtest$cc==1] , unlist(dfnew[1,3]-dfnew[1,2]), unlist(dfnew[1,3]+dfnew[1,2]) ) & between ( newtest$value[newtest$cc==2 ] , unlist(dfnew[2,3]-dfnew[2,2]), unlist(dfnew[2,3]+dfnew[2,2]))
    & between ( newtest$value[newtest$cc==3] , unlist(dfnew[3,3]-dfnew[3,2]), unlist(dfnew[3,3]+dfnew[3,2])))

newtest$value<- rep(tmp$value,3) *newtest$value
tester<- newtest %>% as.cimg
return(list(tester,tmp,dfnew))
}
samplemat3<- indicatorMatrix(plant,900:1000 ,850:950,'RBG')
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values
#Isolate Leaves
plot(samplemat3[[1]])

Correlation fft2

#Correlate2D<-function ( A, B ){
 #   a = ft.fft2(A) 
  ## c = a %*% b
#    C = ft.ifft2( c )
 #   C = ft.fftshift(C)
 #  return C
#}
#bw <- plant %>% grayscale
#ff <- FFT(bw)
#a=ff %>% as.data.frame 
#plot(ff$real,main="Real part of the transform")
samplemathsv<- indicatorMatrix(plant,900:1000 ,850:950,'HSV')
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values
#Isolate Leaves
plot(samplemathsv[[1]])

Erode Data

#Correlate2D<-function ( A, B ){
 #   a = ft.fft2(A) 
  ## c = a %*% b
#    C = ft.ifft2( c )
 #   C = ft.fftshift(C)
 #  return C
#}
#bw <- plant %>% grayscale
#ff <- FFT(bw)
#a=ff %>% as.data.frame 
#plot(ff$real,main="Real part of the transform")
gray<- grayscale(samplemat3[[1]] )
grayshrink<- shrink(gray,8)
#Isolate Leaves
plot(grayshrink)

Grow Data

#Correlate2D<-function ( A, B ){
 #   a = ft.fft2(A) 
  ## c = a %*% b
#    C = ft.ifft2( c )
 #   C = ft.fftshift(C)
 #  return C
#}
#bw <- plant %>% grayscale
#ff <- FFT(bw)
#a=ff %>% as.data.frame 
#plot(ff$real,main="Real part of the transform")
gray<- grayscale(samplemat3[[1]] )
graygrow<- grow(gray,2)
#Isolate Leaves
plot(graygrow)

Eigenvalue

#Correlate2D<-function ( A, B ){
 #   a = ft.fft2(A) 
  ## c = a %*% b
#    C = ft.ifft2( c )
 #   C = ft.fftshift(C)
 #  return C
#}
#bw <- plant %>% grayscale
#ff <- FFT(bw)
#a=ff %>% as.data.frame 
#plot(ff$real,main="Real part of the transform")
gray<- grayscale(plant )
pcamat<- gray %>% as.data.frame %>% data.table
pcamat2<- data.table::dcast(pcamat,x~y, value.var='value') %>% as.matrix
photo.pca <- prcomp(pcamat, center = F)

#PCs <- c(round(0.9 * nrow(pcamat2)))
photo<- pcamat
photo[,3]<- pcamat[,3]*photo.pca$x[3,1]

photo%>% as.cimg%>% plot
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values

#Isolate Leaves

Compute Area

Area<- sum(samplemat3[[2]]$value)
print(Area)
## [1] 102595

Compute Perimeter

binImageShrink <- samplemat3[[2]] %>% as.cimg %>% shrink(1) %>% as.data.frame
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values
Perimeter<- sum(samplemat3[[2]]$value)- sum(binImageShrink$value)# ) %>% as.cimg %>% 

print(Perimeter)
## [1] 102595

Create edge figure

binImageShrink <- samplemat3[[2]] %>% as.cimg %>% shrink(1) %>% as.data.frame
## Warning in as.cimg.data.frame(.): Guessing image dimensions from maximum
## coordinate values
sum(samplemat3[[2]]$value)- sum(binImageShrink$cc)# ) %>% as.cimg %>% plot
## [1] 0
#Isolate Leaves

Mean Point

subsetDF<-  subset (samplemat3[[2]], value==1) 
mean(subsetDF$x)
## [1] 776.762
mean(subsetDF$y)
## [1] 872.7603
#Isolate Leaves

Edge Detection

#Use cannyedges
leafimage<-subim(plant, x%inr% c(450,600),y%inr% c(1175,1500))
edges<-cannyEdges(leafimage) 
## Warning in cannyEdges(leafimage): Running Canny detector on luminance
## channel
plot(edges)

#Isolate Leaves

Try the EBImage package

#Use cannyedges
library('EBImage')
## 
## Attaching package: 'EBImage'
## The following object is masked from 'package:data.table':
## 
##     transpose
## The following objects are masked from 'package:imager':
## 
##     channel, dilate, display, erode, resize, watershed
img<-readImage('~/Documents/Github/PhD//imageprocessing/burseragrav1.jpg')
grayimage<-channel(img,"gray")

crop<-grayimage[450:600,1175:1500]
  y = gblur(crop, 3) < .8
  y <- bwlabel(y)
  contours = ocontour(bwlabel(y))
  c = localCurvature(x=contours[[1]], h=11)
  i = c$curvature >= 0
  pos = neg = array(0, dim(crop))
  pos[c$contour[i,]+1]  = c$curvature[i]
  neg[c$contour[!i,]+1] = -c$curvature[!i]
  display(10*(rgbImage(pos,  neg)), title = "Image curvature")      

#Isolate Leaves